Creating an AI-led newsroom is a journey, but committing to making that journey can benefit the entire company, according to Aviral Mathur, head of data and analytics at India’s HT Media.

Mathur and other media leaders from OVB Media, JP/Politikens Hus, Bayerischer Rundfunk, Nordwest Media, and Mediahuis Nord shared how their companies are using AI to improve internal workflows during the recent Generative AI Master Class, part of INMA’s Generative AI Initiative.

After building AI stacks for different companies over the past 15 years, Mathur helmed HT Media’s AI transformation journey, which includes creating a generative AI-based newsbot, improving personalisation, working on audience monetisation and revenue management, and developing new subscription acquisition and churn models.

“The entire AI initiative has been supported by the leadership team of HT, and we have seen a huge boost in our traffic,” Mathur said. “AI is the strong pillar of this growth that we have seen.”

Mathur outlined three metrics the company developed to measure success, with a focus on efficiency as a primary factor:

1. How many hours it could save its journalists and make them more efficient.
2. How it could improve quality as measured by the amount of time users spent on individual articles.
3. What kind of traffic it generated overall: “HT has about 200 million users who come to our sites every month, so traffic is a very important metric for us,” Mathur said.

OVB Media improves productivity

OVB Media in Germany also saw a clear opportunity to leverage AI and improve workflows in its editorial processes.

The company produces seven local editions six days a week, with most of its content sent in by freelancers in the community. Getting all those stories edited and ready for print was consuming a great amount of resources.

“We thought if we can streamline that process with the help of AI, we can get a lot of acceptance for the idea of an AI tool and generative AI helping us in the newsroom,” Florian Schiller, managing director, said.

That led to developing a tool called Wortwandler, which in English translates to “word changer.”

OVB Media started small, quietly rolling it out to the newsroom and enlisting “a couple of people we thought would be open to that idea.” They started using Wortwandler and had a positive response, so OVB set up a day of AI workshops in the newsroom to unveil the project and explain what they wanted to do with it.

From there, they moved into beta testing in the newsroom and began using it on different editions. Within a couple of weeks, every edition was being produced using the AI tool.

“It massively changed the way they produce their pages,” Schiller said, noting it now takes only 10 minutes to prepare an article for print — something that used to take about two hours. “What we also found out was that we actually increased our quality and the output. We saw that we got less mail complaining about basic things like spelling errors but also quality of articles and wording.

“So in the end it actually improved the overall quality of the whole newspaper. And we made some massive gains in productivity.”

Bayerischer Rundfunk focuses on systems integration

Uli Köppen, the head of the AI + Automation Lab at Bayerischer Rundfunk in Germany, said one significant challenge is the need to adapt AI tools to fit into the traditional and digital workflows that coexist in media organisations.

Many newsrooms still operate with legacy systems, and introducing AI tools requires careful consideration of how these new tools interact with old systems.

"You can’t always use browser tools,” Köppen said. “Sometimes you can use them, but you have to integrate them into existing workflows. And this is the cultural change we have to achieve as an industry.”

This reflects the need for a holistic approach where technology adoption is accompanied by changes in organisational culture and processes, she said, pointing to the importance of system integration: “System integration is one of the hardest nuts we have to crack."

Bayerischer Rundfunk has had successful integration of AI to filter and manage user comments on news platforms. This AI-driven tool not only helps in sorting relevant feedback but is integrated directly into systems like Microsoft Teams, which journalists already use, thus minimising disruption and enhancing adoption.

“The newsroom came to us and asked us to do this,” Köppen said. “So we solved the real pain point with our natural language generation service. But what was really an eye opener for us is that we built an alert system that pumped the alerts directly into a system they are already working with.”

JP/Politikens Hus creates tools for its digital workflow

As the head of AI projects at JP/Politikens Hus and someone who has been with the Danish media company for over a decade, Kasper Lindskow has seen the growth and evolution of AI.

JP/Politikens started more strongly committing to AI in 2020 to develop a “richer” news experience for users that aligns with its values.

In 2023, Lindskow’s team built a set of tools that could be used “in a digital workflow that used Generative AI to transform inputs like article drafts or life blogging.” Journalists could utilise the portions of texts that are generated to add to the news Web site or to help refine their drafts for publishing.

Another internal product, MAGNA Basic, is similar to a chatbot like ChatGPT that allows employees “to interact with a large language model.” This secured chat platform does not transmit data to Open AI, thus safeguarding personal information.

MAGNA Basic is a foundational product that can be built upon products journalists are already using. The future of MAGNA is the integration of an editorial workflow, with a CMS or other tools journalists deem necessary, all while enhancing the news experience. Editors will also have access and user controls.

Ultimately, this focus on the news experience is “to enrich how journalists tell news stories,” to include audiovisual content.

Nordwest Media improves its sales process

Thorge Schramm, senior social media and product manager at Nordwest Media Group in Germany, said as the digital advertising side rapidly grows, it presents greater complexities and challenges for the sales team.

“People were trained in the print and are not very familiar with digital,” Schramm said. “So there is a lack of interest and knowledge. People don’t always ask questions because they feel embarrassed — and this hurts our digital sales. So we need to make our sales team more skilled in digital.”

To do that, it created SalesMate, an automated intelligent tool “that helps us transfer knowledge and actively support sales with information about products, internal expertise, and marketing knowledge,” Schramm said.

This sales support sidekick is always accessible and has extensive knowledge based on a large language model, OpenAI, and ChatGPT, and has been customised to meet Nordwest Media’s requirements. 

Schramm noted the importance of constantly reminding employees about AI.

“We can’t remind the team enough to use AI and simplify processes,” Schramm said. “We hold regular master classes in our media group and workshops for this.”

His team also created an AI lab to familiarise employees with AI and have the chance to work with it: “Everyone has to interact with it. Everyone had fun and hopefully learned something about AI.”

Mediahuis Nord streamlines customer feedback processing

When Mediahuis Noord in the Netherlands saw that compiling, analysing, and interpreting customer feedback was taking team members a lot of time and pulling resources from other places, the news publishers decided to use generative AI to help streamline the process.

Mediahuis Noord collects customer feedback through CSAT, a customer satisfaction score. Customers leave open ended responses in six different categories. Mediahuis Noord found the responses to be both rich in insight and filled with irrelevant information.

“This complicates the extraction of actionable data and it could incur additional costs and processes,” said Pieter de Jong, customer experience project manager at Mediahuis Noord. “It’s not only time consuming and cost heavy, but also can be tedious, repetitive and with less quality over time due to limited time and resources.”

Mediahuis Noord implemented a data dashboard where after the customer fills in a CSAT survey, it gets stored on a server. The responses are extracted and organised using AI and the data is uploaded to the software, PowerBI, where different departments can access it.

Mediahuis Noord has seen a 91.7% reduction in operational costs and an increase in the number of feedback points they can analyse which includes both positive and negative comments.

“We established a direct reporting system to relevant departments with up to 90% reliability,” de Jong said. “It means that the feedback is not only quickly processed but also accurately reflects the customer’s voice. These improvements have not only streamlined the operations but also have enhanced our responsiveness to customer needs, having for a more agile and customer-focused business as well.”

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .